Association between Snoring and High-Risk Carotid Plaque Features

Abstract

Objectives

Previous studies have demonstrated an association between snoring and carotid disease independent of sleep apnea. The aim of this study was to quantify the association between self-reported snoring and high-risk carotid plaque features on magnetic resonance imaging (MRI) that predict stroke.

Study design

Cross-sectional

Setting

Tertiary care university hospital and affiliated county hospital

Methods

We surveyed 133 subjects with asymptomatic carotid artery disease that had been previously evaluated with high-resolution MRI. The survey captured data on self-reported snoring (exposure) and covariates (age, sex, body mass index, and sleep apnea by the STOP-Bang questionnaire). A subset of patients underwent home sleep apnea testing. High-risk carotid plaque features were identified on the high-resolution MRI and included thin/ruptured fibrous cap and intraplaque hemorrhage (outcomes). We quantified the association between snoring and high-risk carotid plaque features with the chi-square test (unadjusted analysis) and multivariate logistic regression adjusting for the covariates.

Results

Sixty-one (46%) of 133 subjects surveyed responded; 32 (52%) reported snoring. Significantly higher proportions of snorers than nonsnorers had a thin/ruptured fibrous cap (56% vs. 25%, p=0.01) and intraplaque hemorrhage (63% vs. 29%, p<0.01). In multivariate analysis, snoring was associated with thin/ruptured fibrous cap (OR 4.4, 95% CI 1.1–16.6, p=0.04) and intraplaque hemorrhage (OR 8.2, 95% CI 2.1–31.6, p<0.01) after adjusting for age, sex, BMI, and sleep apnea.

Conclusion

This pilot study suggests a significant independent association between snoring and high-risk carotid plaque features on MRI. Further study is warranted to confirm these results in a larger cohort of subjects.

Introduction

Carotid atherosclerosis is the primary etiologic factor in ischemic stroke, which is a leading cause of disability and death in the United States ^1,2. Several studies have reported an association between snoring and cerebrovascular pathology ^3–11. Most authors have attributed this finding to the overlap of snoring and obstructive sleep apnea (OSA), which is an independent risk factor for stroke ^12–18.

Some research suggests that snoring may contribute to carotid disease independent of OSA. A recent meta-analysis demonstrated an increased relative risk of stroke in self-reported habitual snorers ¹⁹. After adjusting for cardiovascular risk factors, Lee et al. found that heavy snoring (>50% of the night) was significantly associated with carotid but not femoral atherosclerosis in subjects with mild OSA ²⁰. Li et al. found a significant independent association between self-reported severe snoring (≥5 nights per week) and both carotid thickness and the presence of plaque on ultrasound ²¹. In a population-based study by Olson et al., snorers without OSA had significantly higher odds of occlusive vascular disease than non-snorers ²².

Though there is evidence to support the association between snoring and carotid disease, the mediators of this association are unclear. Carotid atherosclerosis is an insidious process of lipid deposition, endothelial dysfunction and inflammation that results in vessel wall plaque formation and luminal stenosis ²³. Though degree of stenosis is the clinical measure of disease severity ²⁴, the majority of ischemic cerebrovascular events occur in subjects with mild-to-moderate stenosis ²⁵. The majority of strokes arise from sudden rupture of the “vulnerable plaque” with consequent atheroembolization. Plaque vulnerability is not defined by size, but by thinning of the fibrous cap overlying the thrombogenic, lipid-rich necrotic core, often preceded by intraplaque hemorrhage ²⁶. High-resolution magnetic resonance imaging (MRI) has emerged as a tool capable of precisely measuring the carotid plaque composition in order to identify plaques most likely to rupture ^27–29. In a prospective cohort study of subjects with asymptomatic 50% to 79% stenosis, Takaya et al. found that plaques with intraplaque hemorrhage and thin/ruptured fibrous cap were 5 and 17 times more likely to develop subsequent cerebrovascular ischemia, respectively ²⁸.

Plaque rupture results from a combined process of inflammation and biomechanical stress ^30,31. Studies of the impact of biomechanical stressors on vasculature have shown that vibration energy – including that produced by snoring – results in vasoconstriction, sympathetic activation and endothelial damage ^32–38. In addition, vibration trauma may contribute to mechanical stress on the carotid wall, which predicts of plaque rupture ^39,40. Therefore, it is biologically plausible that snoring may be a risk factor for stroke through a mechanism of vibration trauma that is independent of the hypoxia and sympathetic activation that characterizes OSA.

The first aim of this study was to quantify the association between snoring and high-risk carotid plaque features on MRI. We hypothesized that there would be a significant association between self-reported snoring and high-risk carotid plaque as defined by 1) thin or ruptured fibrous cap, and 2) intraplaque hemorrhage, independent of age, sex, body mass index (BMI), and OSA (defined by STOP-Bang score ≥4). In addition, we hypothesized that those who snored more frequently and more loudly would have the highest percentages of high-risk MRI features. The second aim was to pilot objective measurement of OSA in the existing cohort of carotid MRI patients via home sleep apnea testing, in order to assess feasibility of home sleep apnea testing in this retrospective cohort and begin to estimate the objective prevalence and severity of OSA in this cohort.

Methods

Patients and recruitment

Sleep surveys were mailed to 133 subjects who had undergone high-resolution MRI of their carotid arteries. Subjects were part of a longitudinal cohort recruited 1) to identify MRI plaque characteristics predictive of stroke, and 2) to follow the natural history of carotid plaques on imaging ^28,41,42. Subjects with asymptomatic carotid stenosis by duplex ultrasound were recruited from the University of Washington Medical Center (UWMC) and affiliated hospitals. All subjects underwent baseline MRI at the UWMC Vascular Imaging Lab using a previously published protocol ^28,43–45. Images were analyzed by two experienced reviewers blinded to clinical information who identified areas of hemorrhage and categorized fibrous caps as either “thick” versus “thin or ruptured” using published, histologically validated criteria ^43,46,47. Subjects were excluded from the original cohort if they had contraindication to MRI, limited life expectancy, severe chronic illness or chronic disability, age <40 or >80 years, prior carotid surgery or neck radiation, history of transient ischemic attack (TIA) or stroke within the five years preceding the MRI, or pregnancy.

Subjects who had not withdrawn from the original cohort were contacted. Those willing to participate and able to give informed consent were included. This study was approved by the Institutional Review Board of the University of Washington.

Data Collection

Subjects completed the survey on paper or online via REDCap, a secure, web-based research application ⁴⁸. Subjects were encouraged to ask for help from a bed partner if available. Patients were asked to indicate on a Likert scale the frequency and severity of snoring at the time of their baseline study MRI (date of MRI provided as a reference). The sleep survey was administered between August 2013 and May 2015, whereas MRIs occurred between January 2000 and April 2015. Because some MRIs occurred years prior to the date of survey, subjects were also asked to report on their snoring at the present time for use in a secondary analysis. Analysis of present snoring does not preserve the correct temporal relationship between snoring and outcome. However, we theorized that present snoring was likely to be more accurately recalled and could serve as a proxy for past snoring, as snoring is unlikely to change in late adulthood^49–51.

Embedded within the survey were questions required to calculate a STOP-Bang score for each subject. The STOP-Bang is an 8-item validated OSA screening instrument ^52,53. Subjects score one point for each item if they report snoring louder than talking, daytime sleepiness, observed apneas, hypertension, BMI > 35, age > 50, male sex, or neck circumference > 16 inches. A score of ≥4 was chosen a priori to identify subjects likely to have OSA, which corresponds to a 97% positive predictive value for having OSA (apnea hypopopnea index, AHI ≥ 5) ⁵³.

Eighteen subjects (30%) did not report neck circumference, which was required to calculate STOP-Bang score. Twelve subjects had scores too high or too low for this missing variable to impact their classification of OSA at the predetermined cut off and as such they were included in the analysis. Due to the importance of STOP-Bang to the analysis, neck circumference was imputed for the remaining six using published data on the relationship between BMI and neck circumference ⁵⁴. This imputation resulted in an increase in STOP-Bang score from <4 to ≥4 in one subject. STOP-Bang score was not calculated for five subjects due to multiple missing variables.

Demographic and anthropomorphic data including age, sex, height, weight, blood pressure, and health history had been previously collected as part of the MRI cohort study. This occurred via standardized survey and/or direct measurement at the time of MRI.

Sleep testing

In order to pilot sleep testing for OSA in this cohort, a subset of subjects was administered home sleep apnea test using the Watch-PAT^TM. This is a validated, FDA-approved home sleep apnea test device that measures arterial tonography and oximitry to identify apneas and hypopneas⁵⁵. The recording software automatically scores the data and provides a report and hypnogram with AHI (hypopneas include a 4% desaturation) and other sleep testing parameters. AHI was chosen as the primary OSA severity variable.

The recruitment goal was 20 subjects. Survey respondents living within one hour of UWMC were contacted by phone and offered home sleep apnea testing. The test was delivered to the subject’s home and instruction provided by a member of the research team, who collected the device and downloaded the data when testing was complete.

Analysis

Statistical analysis was conducted with Stata/SE 13 software (StataCorp LP, College Station, Texas). Descriptive summaries for continuous variables are reported as means ± standard deviations, whereas frequencies are reported for categorical variables.

Subjects self-reported snoring, specifying frequency and loudness. Frequent snoring was defined as ≥ 5 nights per week ²¹. Loud snoring was defined as snoring “louder than talking”. Pearson’s Chi-squared statistic or the Fisher exact test was used to test the association between any snoring, frequent snoring, and loud snoring with each of the two high-risk MRI plaque characteristics: 1) thin/ruptured fibrous cap and 2) intraplaque hemorrhage.

Multivariate logistic regression with robust standard errors was used to test the association between snoring and plaque characteristic adjusting for age, sex, BMI, and OSA. Subjects with a STOP-Bang score ≥4 were considered to have OSA for the purpose of this analysis.

A separate analysis was performed on the subset of patients in whom objective OSA data were obtained with home sleep apnea testing. Subjects with AHI ≥15 were considered to have OSA. This is a cutoff for moderate or greater OSA, which is associated with increased risk of stroke^15,17,18,56.

Results

Description of sample

Sixty-one (46%) of 133 subjects completed the survey. Forty-seven subjects did not respond, 22 (17%) actively declined, two (2%) were deceased and one (<1%) had advanced dementia.

Respondents were on average 68 years old and mildly overweight (Table 1). The majority was male and hypertensive. Thirty-two respondents (52%) reported snoring at the time of their MRI. Snorers did not differ significantly from nonsnorers with regard to age, sex, time between MRI and survey, BMI, hypertension and past stroke/TIA (Table 1, p>0.05).

Table 1.

Characteristics of respondents overall and by self-reported snoring status (N=61)

	Snoring		All Respondents
	No N=29	Yes N=32	N=61
Variable	Mean (SD) or %	Mean (SD) or %	Mean (SD) or N (%)	Missing N (%)
Age at MRI (years)	68 (8)	68 (8)	68 (8)	0
Body mass index (kg/m2)	25.1 (4.2)	26.6 (4.7)	25.9 (4.5)	0
Time to survey (months)	51 (52)	57 (48)	54 (50)	0
Male	42	58	43 (70)	0
Hypertension	43	57	49 (80)	3 (5)
Past stroke/TIA	30	70	10 (16)	2 (3)
Sleep apnea*	21	79	29 (48)	5 (8)
Help from partner	25	75	24 (39)	0
Frequent snoring**	0	100	8 (13)	4 (7)
Loud snoring**	0	100	7 (12)	3 (5)

TIA = transient ischemic attack; CPAP = continuous positive airway pressure

^*Sleep apnea values include imputed data for five subjects.

^**Frequent snoring defined as ≥ 5 nights/week; loud snoring defined as louder than talking.

Eight (13%) were frequent snorers (≥5 nights/week) and seven (12%) snored louder than talking. Twenty-nine subjects (48%) had OSA defined by a STOP-Bang score ≥ 4. A significantly higher percentage of snorers had OSA (79% vs. 21%, p<0.001). Twenty-four (39%) had help from a bed partner in answering the questions. There was no association between snoring and time between MRI and sleep survey (p > 0.05), but among those who had help from a bed partner, a larger proportion reported snoring than among those who did not have help (75% vs. 38%, p<0.01).

Twenty-five (41%) subjects had a thin/ruptured fibrous cap on MRI whereas 28 (46%) had intraplaque hemorrhage. Mean age was greater for those with thin/ruptured fibrous cap than without (71 vs. 66 years, p=0.03), but did not differ significantly for plaque hemorrhage (p>0.05). Mean BMI and average time between MRI and sleep survey did not differ significantly by either plaque feature (p>0.05).

Association between snoring and MRI plaque features that predict stroke

A significantly greater percentage of snorers than nonsnorers had a thin/ruptured cap and intraplaque hemorrhage (Table 2). When snoring was stratified by frequency, significant differences in MRI features were not seen. Thirty-eight percent of frequent snorers (≥ 5 nights/week) had thin/ruptured cap on MRI; however, this was lower than in those who snored less frequently (Table 2). A similar pattern was observed for loudness, and also when the stratified snoring variables were tested for association with intraplaque hemorrhage (Table 2). Those with high-risk plaque features had higher percentages of OSA than those without, but these differences were not statistically significant (Table 2). This analysis was repeated for self-reported snoring at the time of the sleep survey and the results did not differ from past snoring at the time of MRI (data not shown).

Table 2.

Association between demographic and sleep characteristics of respondents and plaque features that predict stroke (N=61)

	Thin/ruptured Fibrous Cap					Intraplaque Hemorrhage
	No N=36		Yes N=25			No N=33		Yes N=28
Variable	Mean (SD)		Mean (SD)		p	Mean (SD)		Range		p
Age at MRI (years)	66 (1)		71 (2)		0.03	67 (1)		69 (2)		0.4
Body mass index (kg/m2)	25.9 (4.8)		24.0 (5.2)		0.1	25.6 (5.1)		24.4 (5.0)		0.4
Time to survey (months)
Variable	n	%	n	%	p	n	%	n	%	p
Sex
Female	13	72	5	28	0.2	14	78	4	22	0.02
Male	23	53	20	47		19	44	24	56
Hypertension
No	5	56	4	44	0.8	6	66	3	33	0.3
Yes	29	59	20	41		24	49	25	51
Past stroke/TIA
No	32	65	17	35	0.04	29	59	20	41	0.1
Yes	3	30	7	70		3	30	7	70
Snoring
No	22	76	7	24	0.01	21	72	8	28	<0.01
Yes	14	44	18	56		12	38	20	63
Snoring frequency
Nonsnorer	22	76	7	24	0.09	21	72	8	28	0.07
< 5 nights/week	9	45	11	55		8	40	12	60
≥ 5 nights/week	5	62	3	38		4	50	4	50
Snoring loudness
Nonsnorer	22	76	7	24	<0.01	21	72	8	28	0.01
≤ talking	8	36	14	64		7	32	15	68
> talking	6	86	1	14		5	71	2	29
Sleep apnea
No	20	74	7	26	0.08	16	59	11	41	0.6
Yes	15	52	14	48		15	52	14	48

SD = standard deviation MRI = magnetic resonance imaging; TIA = transient ischemic attack; OSA = obstructive sleep apnea;

Multivariate analysis

Multivariate logistic regression analysis was performed to test the association between snoring and each of the two plaque features, adjusting for covariates. Snorers had over four times the odds of having a thin/ruptured cap on MRI than nonsnorers, adjusting for age, sex, BMI and OSA (Table 3). The odds of intraplaque hemorrhage were eight times higher among snorers than nonsnorers, adjusting for age, sex, BMI and OSA (Table 3).

Table 3.

Multivariate logistic regression analysis of the association between self-reported snoring and MRI plaque features that predict stroke (adjusted for age, sex, BMI, and OSA)

	Thin/ruptured Fibrous Cap			Intraplaque Hemorrhage
Variable	OR	95% CI	p	OR	95% CI	p
Snoring	4.2	1.3–12.3	0.01	4.4	1.5–13.0	<0.01
Snoring and age	4.1	1.4–13.1	0.01	4.4	1.5–13.0	<0.01
Snoring and sex	3.8	1.2–11.4	0.02	4.0	1.3–12.3	0.02
Snoring and BMI	4.8	1.5–24.8	0.01	4.9	1.6–14.7	<0.01
Snoring and OSA	4.2	1.1–16.2	0.04	6.1	1.4–26.4	0.01
Snoring, age, sex, BMI, OSA	4.4	1.1–16.6	0.04	8.2	2.1–31.6	<0.01

MRI = magnetic resonance imaging; BMI = body mass index; OSA = obstructive sleep apnea; OR = odds ratio

Home sleep apnea test subset

Of the 31 eligible respondents contacted, 18 (58%) agreed to undergo home sleep apnea testing. Of those, 17 completed testing and one was unable to sleep with the device. Subjects tested did not differ significantly from those not tested with regard to either plaque characteristic, demographics, snoring (overall, loudness or frequency), and STOP-Bang score (all p < 0.05). Those sleep tested had significantly less time between MRI and sleep survey than those not tested (mean 16 vs. 44 months, p<0.01).

Mean AHI was 14 events/hour (SD 13.7, range 2 – 55). Five (29%) subjects had an AHI ≥ 15, indicating moderate or severe OSA. Nine (53%) reported snoring at the time of MRI. One (6%) reported frequent snoring, and two (12%) loud snoring.

When the association between snoring and outcome was tested in the subset, a greater percentage of snorers than nonsnorers had thin/ruptured cap and intraplaque hemorrhage on MRI (Table 4). A higher percentage of those with AHI ≥ 15 had a thin/ruptured cap, but a lower percentage had intraplaque hemorrhage on MRI.

Table 4.

Fisher’s exact test of association between snoring and plaque features that predict stroke within subjects who underwent home sleep apnea testing (N=17)

	Thin/ruptured Fibrous Cap					Intraplaque Hemorrhage
	No N=9		Yes N=8			No N=8		Yes N=9
Variable	n	%	n	%	p	n	%	n	%	p
Snoring
No	5	63	3	37	0.4	5	63	3	37	0.2
Yes	4	44	5	56		3	33	6	67
Sleep apnea (AHI>15)
No	7	58	5	42	0.4	5	42	7	58	0.4
Yes	2	40	3	60		3	60	2	40

AHI = apnea hypopnea index

In order to control for OSA within this small subset, the association between snoring and plaque characteristics was estimated among the 12 subjects without high-risk OSA (all AHI < 15 on home sleep apnea testing). Higher percentages of high-risk plaque features were seen among snorers within this group, but these differences were not statistically significant (Table 5).

Table 5.

Fisher’s exact test of association between snoring and plaque features that predict stroke among subjects with AHI < 15 on home sleep apnea testing (N=12)

	Thin/ruptured Fibrous Cap					Intraplaque Hemorrhage
	No N=7		Yes N=5			No N=5		Yes N=7
Variable	n	%	n	%	p	n	%	n	%	p
Snoring
No	4	67	2	33	0.5	4	67	2	33	0.1
Yes	3	50	3	50		1	17	5	83

AHI = apnea hypopnea index

Discussion

In this sample of subjects with unique carotid phenotyping, self-reported snoring was significantly associated with two high-risk carotid plaque features on MRI that predict stroke. The odds of high-risk features were four to eight times higher in snorers than in nonsnorers of similar age, sex, BMI and OSA classified by STOP-Bang score. We piloted home sleep apnea testing in a subset in order to directly control for OSA. Within this under-powered pilot subset, the association between snoring and high-risk plaque features among those without high-risk OSA were in the same direction as in the overall sample.

We hypothesized that more severe snoring would be associated with high-risk plaque features. However, when snoring was stratified by frequency and loudness, the most severe snoring groups had percentages of high-risk plaque features that were lower than the less severe snoring groups. This finding could be due to misclassification of snoring severity, as snorers may not be able to accurately report how loud or frequently they snore. To evaluate this possibility post-hoc, we limited analysis to those who had help from a bed partner to classify their snoring and found the same overall pattern of results (data not shown).

Unlike the insidious process of lipid deposition and vessel inflammation that forms carotid plaque, hemorrhage and cap rupture are discrete events. There may be a threshold effect of snoring vibration on these specific features rather than a dose-response relationship. It is also possible that snoring severity may not be best quantified by loudness or time spent snoring. A recent study found that increased carotid thickness was associated with snoring at particular sound frequencies, which may be more important to carotid risk than loudness or duration of snoring ⁵⁷.

There was not a significant association between high-risk plaque features and OSA whether defined by STOP-Bang ≥ 4 (within the greater cohort) or AHI ≥ 15 (pilot home sleep apnea test group). This study was not designed or powered to test the association between OSA and stroke; that relationship is well-established ⁵⁸. Small sample size limited the power to detect a significant association between plaque features and OSA within the sleep test pilot subgroup. However, point estimates obtained from this pilot suggest that a larger study may be worthwhile. Within the greater sample, the lack of association may have been due in part to misclassification of OSA. STOP-Bang score has been validated to classify risk of OSA but there is no cutoff to differentiate between mild and moderate OSA. Therefore, any cutoff will include some subjects with mild OSA, which has a weaker or no association with stroke^56,59. Alternatively, cap rupture and intraplaque hemorrhage may be more prone to biomechanical stress (snoring vibration) than to oxidative stress (OSA). Though snoring and OSA impact the carotids by different mechanisms, there has been little investigation into the interaction between the two. A recent prospective study in women showed those with untreated versus CPAP-treated OSA had six times the risk of stroke, which was three fold higher than their risk of coronary heart disease ⁶⁰. Snoring may have been an unmeasured risk factor that contributed to the markedly elevated stroke risk. This is similar to the findings of Lee et al. who showed that heavy snoring was significantly associated with carotid but not femoral atherosclerosis in subjects with OSA ²⁰. The independent effects of treating snoring versus OSA on long-term stroke risk have not been tested.

This study has important limitations. As noted above, measuring snoring by self-report could have introduced exposure misclassification related to the inherent difficulty in recalling snoring at the time of MRI. To assist with recall as much as possible, subjects were provided with the date of their first MRI and were encouraged to ask for help from a bed partner. We could not go back to the time of baseline MRI and measure snoring objectively. Even if this was possible, objective snoring measured during sleep testing is vulnerable to changes in microphone direction, body position and background noise, and only provides a snapshot of a single night ^57,61. Ideally, objective snoring would be measured reliably and include accurate measures of loudness, pitch, energy, and tissue effect. Currently, self- and bed partner-reported snoring is the most common and reliable measure used clinically and in the literature, and gives the best estimate of snoring exposure over time ^{3,4,8,11,62–65}. We collected data on current snoring patterns as a secondary predictor. These data were less prone to recall error, but they rely on the assumption that current snoring reflects past snoring. The overall reported rates of past and present snoring were similar and the results of the analysis of snoring at the two time points did not differ.

While our findings support amassing evidence for a causative relationship between snoring and carotid disease ^{19–21,37,38,65}, other studies have not found a significant relationship between the two^62,66,67. These studies vary in both how they define snoring and carotid disease. This is the first study to examine the relationship of snoring with specific high-resolution phenotyping of carotid disease highly predictive of stroke. The results of this pilot study provide a foundation for a larger-scale prospective study to further examine the relationships between snoring, OSA, carotid disease, and stroke. This has implications for how stroke risk is evaluated and treated. Snoring interventions may ultimately be indicated to prevent the progression of carotid atherosclerosis and stroke.